Matt Bell

In my opinion, it's OK to keep the state in memory since it doesn't have to be huge. For example, after 9 years of the Bitcoin blockchain's existence, the UTXO...

The txs will already be stored in Tendermint so you don't have to worry about that, but if for instance your blockchain was dealing with large data files then it...

*Relevant comment from @davebryson in #15:* >One concern I have is opening rocksdb within Merk. This limits the ability to share the db connection across threads. For example, you may...

>Why is apply_unchecked marked as unsafe? Is this just to alert the user that Batch should be sorted and contain unique keys? Yep, that's correct. If they aren't sorted/unique then...

>why not replace Batch with BtreeMap and remove the additional checks in apply We could possibly make an API for building batches with something like: ```rust db.batch() .put(vec![1], vec![1]) .put(vec![2],...

This gives us the space-saving benefits of a trie without relying on random orderings to balance the tree. We can also use this compact encoding in generated proofs in order...

Hey @ws4charlie, thanks for your interest in merk! As far as I know, updating to `0.15.0` shouldn't cause any issues. I'll be creating a new merk release soon including the...

Awesome, thanks for working on this! And sorry for leaving develop in a half-refactored state, in the future I'll contain half-baked changes in PRs to better communicate what's going on....

BTW, this change will look a lot different after this refactor since it makes some fundamental changes, e.g. supporting concurrency for the backing store. Also, something to keep in mind...

>How fast is fast enough for a Tree in a blockchain environment? You're only as fast as the slowest component - you're still bound by the consensus process and state...